Multiple rewards have asymmetric effects on learning in bumblebees

Animal Behaviour 126 (2017) 123e133

Contents lists available at ScienceDirect

Animal Behaviour journal homepage: www.elsevier.com/locate/anbehav

Multiple rewards have asymmetric effects on learning in bumblebees Felicity Muth a, *, Daniel R. Papaj b, Anne S. Leonard a a b

Department of Biology, University of Nevada, Reno, NV, U.S.A. Department of Ecology and Evolutionary Biology, University of Arizona, AZ, U.S.A.

a r t i c l e i n f o Article history: Received 23 August 2016 Initial acceptance 12 October 2016 Final acceptance 22 December 2016 MS. number: A16-00753R Keywords: Bombus impatiens bumblebee learning memory multitasking nectar pollen pollination

In their natural environments, most animals must learn about multiple kinds of rewards, both within and across contexts. Despite this, the majority of research on animal learning involves a single reward type. For example, bees are an important model system for the study of cognition and its ecological consequences, but nearly all research to date on their learning concerns a single reward, nectar (carbohydrates), even though foragers often simultaneously collect pollen (protein). Features of learning under more ecologically realistic conditions involving multiple reward types are thus largely unexplored. To address this gap, we compared performance on a colour-learning task when floral surrogates offered bumblebees, Bombus impatiens, a single type of floral reward versus multiple, nutritionally distinct rewards. In one experiment, bees learned a floral association with nectar either alone or while simultaneously collecting pollen. In a reciprocal experiment, bees learned a floral association with pollen either alone or while simultaneously collecting nectar. Bees that collected pollen while learning about nectar did not suffer any detriment to learning which flower colour offered nectar. However, this was not the case for the reciprocal task: collecting nectar impaired bees' ability to learn and remember associations between floral colour and pollen. Our findings offer new insight into how bees learn in relation to ecologically realistic rewards and how cognitive constraints may shape their behaviour under ecologically realistic foraging scenarios. © 2017 Published by Elsevier Ltd on behalf of The Association for the Study of Animal Behaviour.

Although research on animal learning often involves a single reward or task type, animals in the wild must often perform multiple tasks, learn multiple associations and correctly remember stimuli across different contexts. These multifaceted demands are generally expected to tax attention, slow acquisition and impair recall. In our own species, cognitive psychology supports the common wisdom that learning proceeds more slowly when subjects are asked to simultaneously perform a second task (Foerde, Poldrack, & Knowlton, 2007; Pashler, 1994; Waldron & Ashby, 2001); more broadly, efficiency and accuracy are often lowered when attention is divided between different activities (Dukas, 2002). For example, blue jays Cyanocitta cristata, detect multiple prey types more slowly compared to a single type (Dukas, 2001), and when bees learn multiple conflicting nectar-foraging tasks, their performance is impaired (Cheng & Wignall, 2006; Chittka & Thomson, 1997). Animals may also suffer impairments to learning when learning about multiple conflicting stimuli in different

* Correspondence: F. Muth, Department of Biology, University of Nevada, Reno, NV 89557, U.S.A. E-mail address: [email protected] (F. Muth).

contexts (e.g. foraging, nest location and oviposition sites; Weiss & Papaj, 2003; Worden, Skemp, & Papaj, 2005; but see Colborn, Ahmad-Annuar, Fauria, & Collett, 1999). However, it is not clear whether these impairments are due to learning a second association generally, or learning a second association that conflicts with the first (i.e. that a given stimulus is rewarding in one context but not in the other). One scenario routinely faced by generalist foragers but rarely explored in research on animal learning concerns learning associations while concurrently collecting multiple resource types. Within this single context (foraging), foragers may encounter prey or diet items that differ in handling techniques and nutritional composition (Simpson & Raubenheimer, 2012). Animals can clearly discriminate between different resources when foraging (Mayntz, 2005; Simpson, Sibly, Lee, Behmer, & Raubenheimer, 2004) and employ different strategies accordingly (Sulikowski & Burke, 2007). Foragers can also learn to associate different stimuli with multiple types of food reward (bees: Muth, Papaj, & Leonard, 2015; locusts: Raubenheimer & Tucker, 1997). However, whether animals incur costs to performance in terms of acquisition or recall when foraging for multiple items is not clear. Understanding the relative costs of learning about a single resource versus multiple resources is

http://dx.doi.org/10.1016/j.anbehav.2017.01.010 0003-3472/© 2017 Published by Elsevier Ltd on behalf of The Association for the Study of Animal Behaviour.

124

F. Muth et al. / Animal Behaviour 126 (2017) 123e133

broadly relevant to understanding the factors that govern the degree of short- or long-term specialization in foraging-related tasks. Generalist bees offer an ideal opportunity to explore how foragers ‘cope’ with learning about multiple reward types. Honeybees (Apis mellifera) and bumblebees (Bombus spp.) visit many different plant species, and in doing so they rapidly form associations between the multimodal ‘sensory billboard’ of the floral display (Leonard & Masek, 2014; Raguso, 2004) and a reward (Leonard, Dornhaus, & Papaj, 2011b). As such, bees have served as important models for our understanding of cognition (Giurfa, 2007; Menzel, 2012; Menzel & Giurfa, 2001). We define cognition as the mechanisms by which animals acquire, process, store and act on information (Healy & Rowe, 2010; Shettleworth, 2010). However, despite the wide range of resources that plants offer pollinators (e.g. resins, oils, oviposition sites, etc.; Armbruster, 2011; Renner, 2006), study of how learning mediates pollination mutualisms has focused almost exclusively on learning in relation to a single type of floral reward, usually nectar (top row in Table 1). In reality, bee-pollinated plants often offer two major nutritionally complementary resources: nectar and pollen (Nicolson, 2011) in diverse combinations (Table 1). In natural settings, therefore, bees encounter learning scenarios considerably more complex than those typical of laboratory-based studies. Recent work has established that bees can learn associations between floral stimuli and pollen rewards (Grüter, Arenas, & Farina, 2008; Muth, Papaj, & Leonard, 2016; Nicholls & Hempel de Ibarra, 2014; Russell, Golden, Leonard, & Papaj, 2015), advancing our understanding of how they learn in relation to non-nectar rewards. Furthermore, bees can learn simultaneously that one floral colour offers only nectar and a second colour only pollen (Muth, Papaj et al., 2015). Yet, despite the fact that individuals of many bee species (including bumblebees) collect both resources on a foraging bout (Goulson, 2003; Hagbery & Nieh, 2012; Hofstede & Sommeijer, 2006; O'Donnell, Reichardt, & Foster, 2000), we know nearly nothing about how learning performance is affected by simultaneously collecting both rewards. Given that bees' learning of floral stimuli has important consequences from both bee and plant

perspectives, our current understanding of learning performance under realistic reward scenarios leaves our picture of the cognitive ecology of pollination surprisingly incomplete. We addressed how learning of a floral feature (colour) was affected when bees foraged for multiple rewards. In a series of freeflying behavioural assays using bumblebees Bombus impatiens as subjects, we compared the learning of a rewardecolour association when bees learned an association with nectar, with or without simultaneously collecting pollen (experiment 1) and when bees learned an association with pollen, with or without simultaneously collecting nectar (experiment 2a). If collecting multiple types of reward impairs learning, then we expected that bees collecting two reward types would find it more difficult to learn an association between a given reward type and a floral feature than bees learning this association in isolation. Alternatively, since individual bumblebees forage for both rewards under natural conditions, and both are critical for colony survival, they may be well equipped to learn associations while concurrently performing these two activities without incurring a cost to task performance. After experiment 2a showed that nectar impaired learning of pollenecolour associations, we explored the mechanism behind this performance decrement in a follow-up experiment (experiment 2b). This experiment tested whether the learning impairment was simply due to carrying out a second task (i.e. collecting nectar) or specifically due to experiencing different rewards on different, conflicting stimuli (Fig. 1). GENERAL METHODS Subjects and Maintenance We used a total of 208 bumblebees from 11 colonies of B. impatiens (Koppert Biological Systems, Howell, MI, U.S.A.) represented equally across treatments within each experiment, with at least 11 subjects per treatment included in the final analysis (for sample sizes see Appendix, Table A1). Colonies were connected to a central foraging arena (L
W
H: 122
59
59 cm) where all

Table 1 Examples of plant strategies for offering nectar and pollen as pollinator rewards Reward

(Nectar

Examples

Reference

Nectar only

Pollen

)

Asclepias Orchidaceae

Pleasants and Chaplin (1983) Sun, Huang, Yu, and Kou (2011)

Pollen only

Pyrola Solanum Dodecatheon Papaver Aster Apiaceae Cucurbitaceae Salix

Knudsen and Olesen (1993) Buchmann (1983) Harder and Barclay (1994) Raine and Chittka (2007) Nisenbaum, Patselas, and Weiner (1999) Langenberger and Davis (2002) Nepi, Guarnieri, and Pacini (2001) Mosquin (1971)

Transition from nectar and pollen to pollen only (protogyny)

Campanula rotundiflora Erythronium grandiflorum

Cresswell and Robertson (1994) Thomson (1986)

Transition from nectar and pollen to nectar only (protandry)

Lavandula stoechas Phacelia linearis Alstroemeria aurea Trachymene incisa

Gonzalez et al. (1995) Eckhart (1991) Aizen and Basilio (1998) Davila and Wardle (2007)

One reward type becomes temporally available

Rhus hirta Sudworth Aralia hispida Lavandula latifolia

Greco, Holland, and Kevan (1996) Thomson, McKenna, and Cruzan (1989) Herrera (1990)

Various combinations of nectar and pollen

The majority of research on bee cognition approximates the ‘nectar only’ reward scenario.

F. Muth et al. / Animal Behaviour 126 (2017) 123e133

125

Experiment 1:

Treatment 1

Treatment 2

Bee learns nectar association, with or without simultaneously collecting pollen

One flower rewarding for nectar, no pollen

One flower rewarding for nectar, pollen on all flowers

Experiment 2a:

Treatment 1

Treatment 2

Bee learns pollen association, with or without simultaneously collecting nectar

One flower rewarding for pollen, no nectar

One flower rewarding for pollen, nectar on all flowers

Experiment 2b:

Treatment 1

Was the impairment to pollen learning in Expt 2a due to interference?

Yellow flowers rewarding for pollen, no nectar

or

or

or

or

Treatment 2 Yellow flowers rewarding for pollen and nectar

Treatment 3 Yellow flowers rewarding for pollen, blue for nectar

Figure 1. Summary of the experiments and treatments. In each case, bees were trained to a colourereward association with a single reward (experiment 1: nectar; experiment 2: pollen) either in the presence or absence of a ‘secondary’ reward (experiment 1: pollen; experiment 2: nectar). Red plus signs (next to circular ‘nectar wells’) indicate nectar present; light grey plus signs (next to rectangular ‘anthers’) indicate pollen present.

training and testing took place. We selected individuals as subjects by allowing a colony access to the test arena containing both a feeder with a cotton wick that offered 50% (w/w) sucrose (hereafter, ‘nectar’) and a chenille stem feeder loaded with ~50 mg of flowercollected cherry pollen (Prunus avium ‘Bing’ variety, Firman Pollen, Yakima, WA, U.S.A.; hereafter ‘Prunus’). Bumblebees are major pollinators of cherry crops (Velthuis & van Doorn, 2006) and readily collect this pollen under laboratory conditions (Muth, Papaj et al., 2015, 2016; Russell & Papaj, 2016). We numbered foragers individually using thorax tags (E. H. Thorne Ltd, Wragby, Lincolnshire, U.K.). We maintained colonies on honeybee-collected pollen (Koppert Biological Systems) prior to experiments, but only used Prunus pollen in experiments. During experiments, bees only had access to the pollen that they collected during training and testing. We maintained colonies on 50% (w/w) scented sucrose (500 ml of water, 500 g of sugar and 2 ml of linalool). By giving bees controlled amounts of nectar and pollen (~4 honeypots filled with 50% sucrose and ~100 mg of pollen at the start of the test day and half their honeypots filled with 50% sucrose and ~700 mg of pollen at the end of the test day), we kept the majority of foragers motivated to collect both nectar and pollen. Floral Arrays and Rewards In all experiments we used arrays of 20 artificial flowers (10 human-blue, 10 human-yellow; for positions in bee colour space see Appendix Fig. A1), each consisting of a coloured corolla and artificial anther, arranged in a 5
4 grid. Flowers (Appendix Fig. A2) were spaced 7 cm apart at the base (5 cm apart at the top) and consisted of three-dimensional printed circular disks, 5 cm in diameter (Makerbot, New York, NY, U.S.A.) placed on inverted plastic tubes (8 cm in height), with a coloured circle (the

‘corolla’, printed on laminated waterproof paper: National Geographic, Washington, D.C., U.S.A.). The centre of the corolla contained two holes: one hole was fitted with a nectar well (diameter: 4 mm), and in the other hole (diameter: 1 mm), an ‘anther’ (a 25 mm beige chenille stem: Creatology, Mountain View, CA, U.S.A.) protruded vertically. Bees readily collect pollen from these artificial anthers (for a description and video see Muth, Papaj et al., 2015, 2016). The design of our artificial flowers made nectar collection and pollen collection behaviour (i.e. probing the well versus contacting the chenille stem) easily distinguishable. Flowers were always oriented such that the nectar wells faced bees as they entered the arena. Between each trial we replaced all nectar wells and anthers. The foraging arena was lit directly from above (59 cm) by both an LED light (2100 lm, Lithonia Lighting, Conyers, GA, U.S.A.) and a human-white LED strip around the top of the inside of the arena (36 W LED, Wholesalers, China), as well as indirectly from fluorescent and natural room light. During training, flowers contained either 12 ml of 50% scented sucrose (for nectar-rewarding flowers) or 12 ml of scented water (for nectar-unrewarding flowers), both scented with 2 ml of linalool per 1000 ml of solution to encourage their discovery. Flowers also either contained ~3e5 mg of Prunus pollen on their anther (for pollen-rewarding flowers) or an empty, scented anther (for pollenunrewarding flowers). We scented unrewarding anthers with pollen (as in Muth, Papaj et al., 2015, 2016) to increase the chance that bees would land on these flowers and to make it more difficult for bees to distinguish flowers based on scent alone. We did this by storing the artificial anthers overnight in a mesh bag within a sealed container of the Prunus pollen, with the bag above the pollen such that they did not come into direct contact. The quantities of nectar and pollen used per flower in the current study have been shown in previous studies to be sufficient to fill the crops and

126

F. Muth et al. / Animal Behaviour 126 (2017) 123e133

pollen baskets of an average-sized forager after collecting the reward from 10 flowers (e.g. nectar: Muth, Keasar, & Dornhaus, 2015; Muth, Papaj et al., 2015; pollen: Muth, Papaj et al., 2015, 2016). General Protocol In all treatments bees had to learn that one of two flower colours predicted a reward, either in isolation or with a second reward also present on flowers (termed the ‘secondary reward’; Fig. 1). In experiment 1 and 2a the secondary reward was present on all the flowers. Bees in all experiments progressed through the following stages: pretraining 1; pretraining 2; training and testing. We carried out both pretrainings on the same day, and training and testing 1e3 days later, also on the same day. The number of days between pretraining and training was represented equally across treatments (mean number of days ± SD: experiment 1: nectar and pollen: 1.4 ± 0.7; nectar only: 1.3 ± 1.0; experiment 2: nectar and pollen: 1.1 ± 0.3; pollen only: 1.3 ± 0.5). Within each experiment, the colour (blue versus yellow) offering the reward was counterbalanced within treatments, across subjects. Pretraining The pretraining stages allowed us to first select foragers that collected both nectar and pollen (~80% of foragers within a colony in the current experiment) and then train them to visit the array used in training and testing phases. Thus, foragers randomly assigned to either the single reward or the dual reward treatments had all previously demonstrated that they collected both nectar and pollen. We did this to ensure that differences between foragers' behaviour were the result of experimental treatments rather than any pre-existing differences among foragers. For pretraining 1, we gave all previously marked foragers access to an array of 20 flowers, identical to those that would be used in training and testing, except that their corollas were grey rather than blue or yellow (Appendix Fig. A1). Each flower contained 36 ml of 50% sucrose and ~10 mg of pollen. Bees that collected both nectar and pollen from these flowers progressed to pretraining 2. In pretraining 2, bees were given access to an array containing 10 blue flowers and 10 yellow flowers. In experiment 1, we gave individual bees (N ¼ 94) flowers that were all rewarding for both nectar and pollen. The pretraining 2 for experiment 2a was conducted as described above, with the exception the flowers for the pollen-only treatment only offered pollen rewards (N ¼ 35) while the nectar and pollen treatment offered both rewards (N ¼ 71). We carried out pretraining in experiment 2a in this way because pilot trials showed that bees would not readily transition from foraging on flowers with both nectar and pollen to pollen-only flowers (often abandoning the array after not finding nectar; an effect not observed when transitioning from dual-reward flowers to nectaronly flowers). However, to rule out the possibility that this difference in pretraining generated the difference in our findings in experiments 1 and 2a, in experiment 2b we carried out pretraining in the same way as experiment 1 (with both nectar and pollen present). The main difference between the results of experiment 1 and 2a was replicated within this single experiment (see Experiment 2b, Results). Training and Testing In all experiments and for each training trial we let an individual bee into the foraging arena with free access to the floral array. During training and testing, a bee could make two kinds of errors (attempting to collect pollen from an unrewarding anther, or

probing an unrewarding nectar well; for more details, see below Behaviour Recorded) and two kinds of correct decisions (collecting pollen from a rewarding anther, or drinking from a rewarding nectar well). A bee was freely allowed to collect nectar and pollen in a trial before returning to the colony. If a bee stopped foraging and did not return to the array she was returned to the colony after 10 min had elapsed. In experiment 1, we gave all bees a minimum of two and maximum of six training trials; when a bee reached a learning criterion of 8 out of 10 sequential correct visits to the flower colour rewarding for nectar (Leonard, Dornhaus, & Papaj, 2011a), we proceeded to the test trial. In experiment 2a, training and testing were identical to experiment 1, with the exception that all bees were given a fixed number of training trials (two). This was because visits to anthers occurred more quickly than visits for nectar, so floral choice for pollen could not be accurately observed live. In previous work (Muth, Papaj et al., 2015, 2016) we found that under these conditions, the vast majority of bees reached the 8 of 10 visits learning criterion within two training trials; review of recordings from experiment 2a established that all bees but one reached this learning criterion. Additionally, while it took bees in experiment 2a fewer trials to learn than bees in experiment 1, bees made many more visits to flowers per trial (cf. Fig. 2a and b): unlike when nectar foraging, bees foraging for pollen returned multiple times to a single flower and successfully extracted pollen from it. After training, we gave all bees a test (probe) trial, where we presented an individual with an array identical to the training array with the exception that all the flowers were unrewarding for nectar and pollen (i.e. containing only scented anthers and scented water). Behaviour Recorded We filmed training and testing using an HD Sony camcorder (30 frames/s) mounted on a tripod placed on top of the test arena facing downwards. From the videos we coded each flower visit made by a bee. This included (1) the colour of the flower (blue or yellow), (2) whether the bee attempted to collect pollen or nectar (i.e. from the anther or nectar well, respectively) and (3) whether she was successful at gaining the reward or not (‘rewarded’ or ‘unrewarded’). Visits to collect nectar were only coded if the bee probed the nectar well with her proboscis, and visits to collect pollen were only coded if the bee made contact with the anther with her antennae or legs. Pollen-collecting behaviours are described in more detail below, as well as in Muth, Papaj, et al. (2015, 2016). Nectar collection was identified as the bee landing on the ‘corolla’ of the flower and probing the nectar well. We defined nectar visits as ‘rewarding’ when the bee drank the nectar reward (probing the well for >2 s; inspection of wells and preliminary observations indicated that bees generally empty the well on their visits, taking 10e30 s to do so) and as ‘unrewarding’ when they probed a well containing water (bees generally left immediately after doing this). If a bee visited a flower and probed for nectar where she had already emptied the well of its reward (but where there was probably some residue), we excluded this from analysis of training, because it is not clear whether bees experience such visits as rewarding or unrewarding, and because they are a different type of error than visiting the CS-. Across all bees' visits to collect nectar, less than 1% were revisits to previously emptied nectar wells. Pollen collection was easily identified as the bee flying directly to the anther and ‘scrabbling’ (Muth, Papaj et al., 2015, 2016) at the pollen with her legs in a stereotyped manner, as in Muth, Papaj et al. (2016; see their Supplementary Video). In packing the pollen onto their corbiculae, bees were also seen extending their

F. Muth et al. / Animal Behaviour 126 (2017) 123e133

6

4

4

2

2 – 31

–

30

20 21

– 11

– 1

Pollen only Pollen and nectar

1 – 1 1 10 – 2 1 20 – 3 1 30 – 4 41 0 – 5 1 50 – 6 1 60 – 7 1 70 – 80

6

Te st

8

40

8

10

Mean number of correct visits (±SE)

(b) 10

Nectar only Nectar and pollen

Te st

(a) 10

127

Visit number Figure 2. Results from the test phase. (a) Experiment 1: the number of correct visits when bees landed and probed for nectar, averaged across blocks of 10 visits when bees either foraged for nectar alone (N ¼ 28) or concurrently collected pollen from both flowers (N ¼ 29). (b) Experiment 2a: the number of correct visits when bees searched for pollen on anthers, averaged across blocks of 10 visits when bees either foraged for pollen alone (N ¼ 30) or concurrently collected nectar from both flowers (N ¼ 24).

proboscis and grooming with their front legs. In our assay, this usually occurred on the flower's anther or in flight, but occasionally the bee would land to groom elsewhere (on a flower's corolla or the arena wall). Because bees only collected and searched for pollen from the flowers' anthers, we defined a ‘visit’ to a flower type as the bee touching the anther with its antennae or legs (either by landing or hovering in front of the top of the anther). If the anther contained pollen and the bee was seen to scrabble with its legs, the visit was recorded as ‘rewarded’. If the anther had no pollen on it and the bee touched it with its antennae or legs, then the visit was recorded as ‘unrewarded’. If the bee briefly touched the rewarding anther with her antennae or legs but did not scrabble to collect pollen from it, then this visit was excluded from analyses because it was not clear whether the bee had failed to detect the pollen or had detected its presence but decided not to collect it (i.e. whether this visit was perceived as reinforcing or inhibiting to the bee). Unlike when foraging for nectar, pollenforaging bees did not empty anthers containing pollen on a single visit. Instead, a bee would generally collect pollen from all rewarding anthers, then return to anthers they had already visited to collect more pollen; these visits were also scored as ‘rewarding’. Across all bees' visits to collect pollen, 58% were rewarding, 28% were unrewarding and 14% were to rewarding flowers but without the bee collecting pollen.

Bees that did not reach the learning criterion within the maximum number of training trials (experiment 1: nectar-only: N ¼ 1; nectar and pollen: N ¼ 3; experiment 2a: nectar and pollen: N ¼ 1) were still included in the learning analysis as they all showed a statistically significant improvement in their proportion of visits to the rewarding flower type (i.e. learning). For each learning analysis we fitted a generalized linear mixed model (GLMER) with a Poisson error distribution using the response variable (number of correct visits within a block of 10 consecutive visits), with the explanatory factors treatment (single reward or two rewards), colour of the reward being trained to (blue or yellow), the continuous variable visit block number (blocked in groups of 10) and the random factors bee nested within colony (‘Model 1’). We measured test performance as the number of correct visits in the first 10 visits a bee made and compared treatments with unpaired t tests. We carried out all analyses in R v.2.15.1 (R Development Core Team, 2010). GLMMs were carried out using the ‘glmer()’ function in ‘lme4’ package (Bates, Maechler, Bolker, & Walker, 2015) and LMMs were carried out using the ‘lme()’ function in ‘nlme’ package, specifying type III sum of squares and sum contrasts in cases where there were interactions (Pinheiro, Bates, DebRoy, Sarkar, & R Core Team, 2016). For all models, maximal models were run initially, and then nonsignificant interactions were removed in a stepwise fashion. In all cases we verified that model assumptions were met.

Data Analysis Ethical Note To compare learning across treatments (single reward versus two rewards, separately for all three experiments), we used the number of correct visits within a block of 10 consecutive visits as the response variable. Because different bees within treatments made a different total number of visits to flowers, we set the upper value of visits at the block where there were fewer than seven bees still visiting flowers (i.e. experiment 1: 4 blocks, experiment 2a, 8 blocks). We excluded bees that did not make at least 15 visits to the trained reward type (experiment 1: N ¼ 3; experiment 2a: N ¼ 7) and bees that did not collect at least two rewards of the secondary reward type during the time they took to visit 20 flowers for the trained reward type (experiment 1: N ¼ 0; experiment 2a: N ¼ 3).

We maintained ethical standards throughout testing and euthanized bees via freezing. RESULTS Experiment 1: Learning Nectar Associations While Concurrently Collecting Pollen Bees' ability to learn a nectarecolour association was not impaired by concurrently collecting pollen. Bees in both treatments learned that one colour predicted a nectar reward, making fewer

128

F. Muth et al. / Animal Behaviour 126 (2017) 123e133

visits to the alternative colour to probe for nectar across visits (Poisson GLMM: z ¼ 4.814, P < 0.0001; Fig. 2a) and the number of errors during learning did not differ between treatments (z ¼ 2.654, P ¼ 0.107). Blue-trained bees made more correct choices than yellow-trained bees across both reward treatments (z ¼ 1.613, P < 0.01). There was no difference between treatments in the number of correct choices bees made in the unrewarded test phase (unpaired t test: t48 ¼ 0.407, P ¼ 0.686; Fig. 2a). Experiment 2a: Learning Pollen Associations While Concurrently Collecting Nectar In contrast to results of experiment 1, bees' ability to learn a pollenecolour association was impaired when they concurrently collected nectar. In a pollenecolour learning task, bees in each of two training conditions learned the flower colour associated with a pollen reward, making fewer errors (i.e. searching for pollen on unrewarding anthers) across visits (Poisson GLMM: z ¼ 5.22, P < 0.0001; Fig. 2b). However, bees that collected nectar while learning this colour association with pollen made more errors while learning than bees that foraged on flowers with no nectar (z ¼ 4.66, P < 0.0001). As in experiment 1, blue-trained bees made more correct choices than yellow-trained bees across both reward treatments (z ¼ 3.42, P < 0.0001). Bees that collected both nectar and pollen performed worse during the unrewarded test phase than bees that foraged for nectar alone (unpaired t test: t43 ¼ 3.36, P < 0.005; Fig. 2b). Motivation to Collect Rewards across Treatments Because we did not control the amount of each reward bees collected, we were interested in whether any differences in learning performance might reflect treatment-related differences in bees' motivation to collect particular rewards. When we compared bees' attempts to collect nectar and pollen in experiments 1 and 2a, we found that the presence of pollen on flowers did not lessen bees' motivation to collect nectar, whereas the presence of nectar on flowers appeared to decrease bees' motivation to collect pollen. In experiment 1, bees made the same number of attempts to collect nectar, regardless of whether they foraged for nectar alone or simultaneously foraged for pollen (unpaired t test: t59 ¼ 0.023, P ¼ 0.982; Fig. 3). In contrast, in experiment 2a, bees

Because we were interested in why bees that collected nectar were impaired in their ability to learn about pollen, we conducted a follow-up experiment to determine whether this impairment was caused by motivational changes associated with the mere act of collecting nectar, or whether it was driven more specifically by collecting nectar on the colour that was unrewarding for pollen (i.e. via interference; Weiss & Papaj, 2003; Worden et al., 2005). We trained bees that yellow flowers contained pollen either (1) when no flowers offered nectar, (2) when yellow flowers also offered nectar, or (3) when blue flowers offered nectar (Fig. 1). If collecting nectar makes bees less motivated to collect pollen and more motivated to collect nectar, we expected a detriment to learning in either treatment where bees collected nectar (regardless of which colour of flower they collected it from) relative to bees

(a)

250

60

40

20

0

EXPERIMENT 2B

Mean number of atempts to collect Pollen across all trials (+SE)

Mean number of atempts to collect nectar across all trials (+SE)

80

made fewer attempts to forage for pollen when they simultaneously foraged for nectar compared to the bees that foraged for pollen alone (unpaired t test: t62 ¼ 4.411, P < 0.0001; Fig. 3). A related analysis regarding how bees' collection of nectar or pollen changed over successive visits differentially based on treatment returned a similar finding. Again, we found that bees' foraging seemed to be driven more by nectar than by pollen, while in general bees preferred to forage on flowers that contained both types of reward. In experiment 1, where all flowers contained pollen, bees made fewer visits to collect pollen across successive trials and were more likely to collect pollen from the flower colour that was also rewarding for nectar (LMMs: experiment 1: blue-trained: colour: F1,83 ¼ 13.813, P < 0.001; trial: F1,83 ¼ 9.603, P < 0.005; yellowtrained: colour: F1,105 ¼ 12.364, P < 0.001; trial: F1,105 ¼ 14.604, P < 0.001). However, the same relationship was not true in experiment 2a, when all the flowers contained nectar. In experiment 2a, blue-trained bees collected more nectar over successive visits and were also more likely to collect nectar from the colour of flower also rewarding for pollen (i.e. blue flowers) (LMMs: visit (blocked): F1,89 ¼ 5.698, P < 0.05; colour: F1,89 ¼ 29.968, P < 0.0001). Yellowtrained bees did not collect more (or less) nectar across visits or from a particular colour overall, but they did collect more nectar from yellow flowers (rewarding for pollen) relative to blue on later visits (colour
visit interaction: F1,95 ¼ 7.948, P < 0.01; visit: F1,95 ¼ 0.009, P ¼ 0.925; colour: F1,95 ¼ 1.590, P ¼ 0.211).

Nectar only

Nectar and pollen

(b)

200

150

100

50

0

Pollen only

Nectar and pollen

Figure 3. Mean number of attempts that bees made during training to collect (a) nectar (experiment 1) or (b) pollen (experiment 2a) either when collecting that reward alone or when collecting both rewards.

F. Muth et al. / Animal Behaviour 126 (2017) 123e133

Methods All bees (N ¼ 71) were pretrained as described above, on arrays containing both nectar and pollen. Training and testing was carried out in the same way as experiment 2a, with all bees given two training trials and one test trial. We trained bees that yellow flowers contained pollen either (1) when no flowers offered nectar (N ¼ 18), (2) when yellow flowers also offered nectar (N ¼ 22), or (3) when blue flowers offered nectar (N ¼ 23) (Fig. 1). We only trained bees that yellow flowers were rewarding for pollen (and not blue) because the largest effect we found in experiment 2a was for yellow-trained bees (the less preferred colour). To determine whether bees made more visits to the rewarding flower type across successive visits and whether this differed between treatments, we fitted a GLMM using the same variables as ‘Model 1’ (see Data Analysis above), except this time the factor ‘treatment’ had three levels and there was no factor of ‘colour of reward being trained to’ since all bees were trained to pollen on yellow flowers. To determine whether the three treatments differed from each other in their test performance, we fitted a linear mixed model (LMM) with the response ‘number of correct choices’, the explanatory factor ‘treatment’ (three levels) and the random factor ‘colony’. Tukey post hoc tests were carried out using the ‘lsmeans’ package. Results Why did nectar impair learning of pollenecolour associations? We found that the mere act of collecting nectar, regardless of whether it was on the same or a different flower as pollen, impaired bees' ability to learn the pollenecolour association (Fig. 4). This finding is consistent with the hypothesis that nectar collection impairs performance in pollen learning by affecting motivation. Across all treatments, bees chose more ‘correct’ flowers rewarding for pollen across successive visits than incorrect flowers (Poisson GLMM: z ¼ 2.287, P < 0.05). However, their performance differed between treatments, with bees in the pollen-only treatment performing better than bees in the other two pollenenectar treatments (no difference between the two nectar treatments: z ¼ 0.436, P ¼ 0.663; a significant difference between pollen-only and other two treatments: z ¼ 2.227, P < 0.05). The test phase showed the same pattern of results: bees in the pollen-only group performed better than the other two treatments (LMM: F2,33 ¼ 3.467, P < 0.05; Fig. 4). GENERAL DISCUSSION Learning is an essential feature of the interaction between plants and their pollinators, yet the study of its mechanisms and consequences usually involves a single reward type scenario (Giurfa, 2007; Leonard & Masek, 2014). This set-up fails to capture the complexity of bees' real-life foraging situations, which often also involves collection of pollen (Table 1). If studies of pollinator cognition are to be relevant for research into the ecology and evolution of plantepollinator relationships, it is critical to understand how bees cope with learning associations in relation to

10

Mean number of correct visits (±SE)

that foraged for pollen alone. Alternately, it is possible that collecting nectar on the flower that was unrewarding for pollen may have interfered with a bee's ability to learn about pollen (as in Worden et al., 2005): a bee remembered a blue flower as rewarding but misremembered which reward she found there, so mistakenly searched for pollen instead of nectar on blue flowers. If so, we expected that only the bees in the treatment where nectar was offered on blue flowers would show impaired learning relative to the two other treatment groups.

129

8

6

4

2

0

Pollen only

Yellow

Blue

Colour of flower rewarding for nectar Figure 4. Experiment 2b: the mean number of correct visits to anthers (of yellow flowers) to search for pollen during the test phase, when bees either foraged for pollen alone on yellow flowers (N ¼ 16) or concurrently collected nectar from either blue flowers (N ¼ 11) or yellow flowers (N ¼ 13).

multiple rewards. Beyond providing insights into how nutritionally complex rewards mediate interactions between plants and their pollinators, these questions are also relevant to any foraging scenario in which an animal must learn multiple associations for nutritionally distinct resources. Here we present the first evidence that foraging simultaneously for multiple floral rewards has asymmetric effects on bee learning. Bees that collected pollen while also foraging for nectar did not suffer any detriment to their ability to learn which flowers offered nectar. This finding supports the assumption that using nectar-only floral surrogates (the approach used in the majority of bee cognition research) produces reasonable estimates of nectar-based learning ability under more realistic conditions. However, this was not the case for the reciprocal task: bees learning a pollenecolour association made more errors (searching for pollen on unrewarding anthers) when they simultaneously collected nectar. The unexpected asymmetry in cognitive costs associated with foraging for two rewards at once raises new questions about how bees handle learning multiple associations, with consequences for understanding the evolution of floral reward complexity. From the pollinator perspective, our results uncover a constraint on bees' ability to learn when foraging for multiple rewards, although only in a single direction. In other systems, carrying out multiple tasks or task switching can lead to lower performance (Cheng & Wignall, 2006; Courage, Bakhtiar, Fitzpatrick, Kenny, & Brandeau, 2015; Monsell, 2003). A detriment to learning is also found when insects learn associations with the same conditioned stimuli in two different contexts (bees: Worden et al., 2005; butterflies: Weiss & Papaj, 2003) or learn two tasks with conflicting response requirements (bees: Cheng & Wignall, 2006; Chittka & Thomson, 1997). In these cases, it seems that bees suffer from interference between the two tasks. However, in our experiment, collecting (and potentially learning about) two rewards simultaneously was not sufficient to cause any detriment to learning, since bees were only impaired when learning about pollen while simultaneously collecting nectar. Additionally, the detriment to learning was not simply caused by bees' confusion over which reward a

130

F. Muth et al. / Animal Behaviour 126 (2017) 123e133

given conditioned stimulus contained (as in Weiss & Papaj, 2003). For example, it was not the case that bees only suffered an impairment to learning that pollen was found on yellow flowers when nectar was found on blue, because bees suffered the same impairment to learning even when they simultaneously collected nectar from the same colour that was rewarding for pollen (experiment 2b). This implies that the learning impairment was caused by collecting nectar per se, regardless of the colour of flower bees collected it from. Combined with our finding that collection of pollen did not impair nectar learning (experiment 1), it seems that, rather than a general cost to learning about multiple reward types, there is something specific regarding nectar that impairs bees' ability to learn about pollen. This asymmetry in nectar and pollen's effects on learning about the other implies that nectar is a more salient reward to bees than pollen, being more immediately motivating. This is further supported by our finding that bees collected less pollen when also foraging for nectar (experiment 2a), but they did not change their effort to collect nectar when also foraging for pollen (experiment 1). Furthermore, motivation to collect pollen waned over time in experiment 1 whereas bees remained motivated to collect nectar in experiment 2a (see Results above). When collecting both rewards at the same time, a higher inherent motivation to collect nectar could be driven by bees' immediate need for carbohydrates to fuel costly flight and foraging (Goulson, 2003; Heinrich, 2004), in contrast to pollen, which infers relatively long-term benefits to offspring survival and colony growth. While the mechanism by which nectar acts as a reward to bees has been established (Hammer & Menzel, 1995), precisely how pollen rewards bees is not clear, although both the taste (Muth, Francis, & Leonard, 2016; Ruedenauer, Spaethe, & Leonhardt, 2015) and nutritional content (Vaudo, Patch, Mortensen, Tooker, & Grozinger, 2016) are important factors in determining bumblebee floral preferences and may influence learning. Elucidating the mechanism by which pollen reinforces bee behaviour will be critical in determining precisely how bees integrate information about nectar and pollen rewards when foraging for both. An obvious next step will be to determine whether these results hold under a range of different nectar and pollen qualities, and whether our findings change with colony nutritional state. However, with respect to the current experiments, we kept colonies maintained on diets restricted for both reward types in an attempt to offer rewards of equal value to bees. This manipulation kept the majority of bees motivated to collect both rewards. Furthermore, the nectar and pollen used seemed to reinforce learning to the same extent when they were collected in isolation, supported by our finding that nectar-only bees in experiment 1 and pollen-only bees in experiment 2a had similar rates of learning (compare single-reward learning curves in Fig. 2a and b). The asymmetry we found is perhaps analogous to a previous finding in honeybees (A. mellifera) that either learned two associations (one appetitive, one aversive) or a single association (Vergoz, Roussel, Sandoz, & Giurfa, 2007). Bees that learned both associations did not suffer any impairment to learning relative to bees that learned a single aversive association (odour þ shock), but were impaired relative to bees that learned a single appetitive association (odour þ sucrose). The impairment to appetitive learning was specifically due to learning the aversive association rather than simply being shocked: bees that were shocked in a protocol where the odour stimulus was unpaired from the shock did not suffer the impairment to appetitive learning. Therefore, just as the bee brain may prioritize learning aversive shock associations over appetitive sucrose associations (Vergoz et al., 2007), so too the current experiments suggests that nectar learning may be prioritized over pollen learning.

Our finding that when flowers offered nectar, bees found it more difficult to learn about floral pollen (experiment 2b), also has implications for understanding the functional ecology of floral rewards from a plant's perspective. While a pollinator's aptitude for learning floral associations is often considered from the individual's or colony's perspective (Raine & Chittka, 2008), bees' learning performance is often equally as significant from the plant's perspective, since pollinators that make more errors while learning an association or show poorer recall of an association are more likely to transfer pollen to heterospecific plants (Chittka & Thomson, 1997). While floral reward diversity has been the subject of a number of overviews (Armbruster, 2011; Renner, 2006), functional explanations for why any one plant might benefit from offering more than one reward type have been relatively unexplored. Any reward strategy is likely to be shaped by multiple agents, and previous explanations have highlighted the potential benefits of increased pollinator diversity (Ghazoul, 2006) or meeting the multinutrient target of a given pollinator (Francis, Muth, Papaj, & Leonard, 2016). Our findings complement and extend these ideas, showing that nectar reduces bees' ability to learn effectively about collecting pollen and makes bees less motivated to collect pollen overall. Since nectar is thought to be a relatively ‘cheap’ reward for the plant compared to reproductively costly pollen (Hargreaves, Harder, & Johnson, 2009), it would presumably benefit the flower to encourage learning of its features by inducing the bee to collect nectar rather than pollen. While it is necessary for plants to encourage some pollen collection for successful reproduction, restriction of pollen collection by a single individual is evident from other mechanisms such as dosing via anther morphology (e.g. Harder & Thomson, 1989), pollen architecture (Lunau, Piorek, Krohn, & Pacini, 2015) and pollen chemistry (Muth, Francis, et al., 2016). The idea that plants benefit from the cognitive limitations of their pollinators is not a novel one: floral constancy, the tendency of individuals to sequentially visit one flower type while bypassing others, may also be explained by cognitive constraints on a pollinator's ability to recognize or handle multiple types (Chittka, Thomson, & Waser, 1999; Dukas, 1998; Goulson, 2000; Lewis, 1993; Waser, 1983, 1986). Likewise, sexually or food-deceptive species may benefit not only from pollinators' innate attractions to elements of their display, but also from difficulty in learning to , Johnson, & Kindlmann, 2006; Pohl, Watolla, avoid them (Jers akova & Lunau, 2008). Such a constraint on bees' ability to learn effectively about pollen associations when collecting nectar could in turn feed back onto the evolution of bee foraging strategies: if bees are constrained in their ability to learn about multiple rewards, this could help explain the evolution of nectar- and pollen-foraging specialists in large social colonies such as honeybees (Page, Scheiner, Erber, & Amdam, 2006). Similarly, it may explain why individuals of some species (e.g. bumblebee species) often specialize on one reward in the short term (Francis et al., 2016), only generalizing when necessary (Hagbery & Nieh, 2012). This is supported by findings from three wild Bombus species that were less efficient when switching from pollen to nectar collection (Cartar, 1992) than when collecting a single nutrient. Thus there may be a trade-off between the cognitive cost of collecting both rewards and the energetic advantage of doing so, especially when limited by the number of foragers in a colony. The complex interplay between plants and their pollinators offers a useful system in which to address further more general questions of how multiple, different rewards that differ in quality can mediate learning. In this system, individual reward types are not only tractable experimentally, but the pattern of natural diversity in plant reward strategies (Table 1) offers reward scenarios in which each type is or is not offered alongside the other. Analogous scenarios might play out in other contexts, for example a

F. Muth et al. / Animal Behaviour 126 (2017) 123e133

generalist predator sampling prey items that vary in nutritional content (Mayntz, 2005), and learning aspects of their antipredator displays, or a seed-disperser learning information about the location and quality of fruits while interspersing collection of insect prey. In each of these cases, the speed and accuracy of learning will have ecological effects: in understanding these effects, our results highlight the need to consider the possibility that multiple ‘rewards’ can have different and interactive effects on cognitive performance. Animals' cognitive abilities are expected to evolve in response to selection (Dukas, 2008), but by experimentally investigating learning with a single reward, we may not be gaining a realistic picture of the ecological relevance of learning. Acknowledgments We thank J. Francis and D. Picklum for comments and Firman Pollen for the pollen. Amanda Scampini, Harvi Singh, Trent Cooper and Rene Bonilla assisted with data collection. Thanks to D. Sowers for creating a web application to assist with video coding. This research was supported by the National Science Foundation (IOS1257762). References Aizen, M., & Basilio, A. (1998). Sex differential nectar secretion in protandrous Alstroemeria aurea (Alstroemeriaceae): Is production altered by pollen removal and receipt? American Journal of Botany, 85, 245e252. Armbruster, W. (2011). Evolution and ecological implications of ‘specialized’ pollinator rewards. In S. Patiny (Ed.), Evolution of plantepollinator relationships (pp. 44e67). Cambridge, U.K.: Cambridge University Press. Bates, D., Maechler, M., Bolker, B., & Walker, S. (2015). Fitting linear mixed-effects models using lme4. Journal of Statistical Software, 67(1), 1e48. http://dx.doi.org/ 10.18637/jss.v067.i01. Buchmann, S. L. (1983). Buzz pollination in angiosperms. In C. E. Jones, & R. J. Little (Eds.), Handbook of experimental pollination biology (pp. 73e113). New York, NY: Van Nostrand Reinhold. Cartar, R. V. (1992). Adjustment of foraging effort and task switching in energymanipulated wild bumblebee colonies. Animal Behaviour, 44, 75e87. http:// dx.doi.org/10.1016/S0003-3472(05)80757-2. Cheng, K., & Wignall, A. E. (2006). Honeybees (Apis mellifera) holding on to memories: Response competition causes retroactive interference effects. Animal Cognition, 9(2), 141e150. http://dx.doi.org/10.1007/s10071-005-0012-5. Chittka, L. (1992). The colour hexagon: A chromaticity diagram based on photoreceptor excitations as a generalized representation of colour opponency. Journal of Comparative Physiology A, 170, 533e543. Chittka, L., & Thomson, J. D. (1997). Sensori-motor learning and its relevance for task specialization in bumble bees. Behavioral Ecology and Sociobiology, 41(6), 385e398. http://dx.doi.org/10.1007/s002650050400. Chittka, L., Thomson, J. D., & Waser, N. M. (1999). Flower constancy, insect psychology, and plant evolution. Naturwissenschaften, 86(8), 361e377. http:// dx.doi.org/10.1007/s001140050636. Colborn, M., Ahmad-Annuar, A., Fauria, K., & Collett, T. S. (1999). Contextual modulation of visuomotor associations in bumble-bees (Bombus terrestris). Proceedings of the Royal Society B: Biological Sciences, 266(1436), 2413. http:// dx.doi.org/10.1098/rspb.1999.0940. Courage, M. L., Bakhtiar, A., Fitzpatrick, C., Kenny, S., & Brandeau, K. (2015). Growing up multitasking: The costs and benefits for cognitive development. Developmental Review, 35, 5e41. http://dx.doi.org/10.1016/j.dr.2014.12.002. Cresswell, J., & Robertson, A. (1994). Discrimination by pollen-collecting bumblebees among differentially rewarding flowers of an alpine wildflower, Campanula rotundifolia (Campanulaceae). Oikos, 69(2), 304e308. Retrieved from http://www.jstor.org/stable/3546151. Davila, Y. C., & Wardle, G. M. (2007). Bee boys and fly girls: Do pollinators prefer male or female umbels in protandrous parsnip, Trachymene incisa (Apiaceae)? Austral Ecology, 32(7), 798e807. http://dx.doi.org/10.1111/j.1442-9993.2007.01757.x. Dukas, R. (1998). Cognitive ecology: The evolutionary ecology of information processing and decision making. Chicago, IL: University of Chicago Press. Dukas, R. (2001). Limited attention: The constraint underlying search image. Behavioral Ecology, 12(2), 192e199. http://dx.doi.org/10.1093/beheco/12.2.192. Dukas, R. (2002). Behavioural and ecological consequences of limited attention. Philosophical Transactions of the Royal Society B: Biological Sciences, 357(1427), 1539e1547. http://dx.doi.org/10.1098/rstb.2002.1063. Dukas, R. (2008). Evolutionary biology of insect learning. Annual Review of Entomology, 53(1), 145e160. http://dx.doi.org/10.1146/annurev.ento.53.103106.093343. Eckhart, V. M. (1991). The effects of floral display on pollinator visitation vary among populations of Phacelia linearis (Hydrophyllaceae). Evolutionary Ecology, 5(4), 370e384. http://dx.doi.org/10.1007/BF02214154.

131

Foerde, K., Poldrack, R., & Knowlton, B. (2007). Secondary-task effects on classification learning. Memory & Cognition, 35, 864e874. Francis, J. S., Muth, F., Papaj, D. R., & Leonard, A. S. (2016). Nutritional complexity and the structure of bee foraging bouts. Behavioral Ecology, 27(3), 903e911. http://dx.doi.org/10.1093/beheco/arv229. Ghazoul, J. (2006). Floral diversity and the facilitation of pollination. Journal of Ecology, 94(2), 295e304. http://dx.doi.org/10.1111/j.1365-2745.2006.01098.x. Giurfa, M. (2007). Behavioral and neural analysis of associative learning in the honeybee: A taste from the magic well. Journal of Comparative Physiology A: Neuroethology, Sensory, Neural, and Behavioral Physiology, 193(8), 801e824. http://dx.doi.org/10.1007/s00359-007-0235-9. Gonzalez, A., Rowe, C. L., Weeks, P. J., Whittle, D., Gilbert, F. S., & Barnard, C. J. (1995). Flower choice by honey bees (Apis mellifera L.): Sex-phase of flowers and preferences among nectar and pollen foragers. Oecologia, 101(2), 258e264. http://dx.doi.org/10.1007/BF00317292. Goulson, D. (2000). Are insects flower constant because they use search images to find flowers? Oikos, 88(3), 547e552. http://dx.doi.org/10.1034/j.16000706.2000.880311.x. Goulson, D. (2003). Bumblebees: Their behaviour and ecology. Oxford, U.K.: Oxford University Press. Greco, C. F., Holland, D., & Kevan, P. G. (1996). Foraging behaviour of honey bees (Apis mellifera L.) on staghorn sumac [Rhus hirta Sudworth (ex-typhina L.)]: Differences and dioecy. Canadian Entomologist, 128(3), 355e366. http:// dx.doi.org/10.4039/Ent128355-3. Grüter, C., Arenas, A., & Farina, W. M. (2008). Does pollen function as a reward for honeybees in associative learning? Insectes Sociaux, 55(4), 425e427. http:// dx.doi.org/10.1007/s00040-008-1022-5. Hagbery, J., & Nieh, J. (2012). Individual lifetime pollen and nectar foraging preferences in bumble bees. Naturwissenschaften, 99, 821e832. Hammer, M., & Menzel, R. (1995). Learning and memory in the honeybee. Journal of Neuroscience, 15, 1617e1630. Retrieved from http://www.jneurosci.org/content/ 15/3/1617.short. Harder, L. D., & Barclay, R. M. R. (1994). The functional significance of poricidal anthers and buzz pollination: Controlled pollen removal from Dodecatheon. Functional Ecology, 8(4), 509e517. http://dx.doi.org/10.2307/2390076. Harder, L. D., & Thomson, J. D. (1989). Evolutionary options for maximizing pollen dispersal of animal-pollinated plants. American Naturalist, 133(3), 323e344. Hargreaves, A., Harder, L., & Johnson, S. (2009). Consumptive emasculation: The ecological and evolutionary consequences of pollen theft. Biological Reviews, 84, 259e276. http://dx.doi.org/10.1111/j.1469-185X.2008.00074.x. Healy, S., & Rowe, C. (2010). Information processing: The ecology and evolution of cognitive abilities. In D. Westneat, & C. Fox (Eds.), Evolutionary behavioral ecology (pp. 162e176). Oxford, U.K.: Oxford University Press. Heinrich, B. (2004). Bumblebee economics. Cambridge, MA: Harvard University Press. Herrera, C. (1990). Brood size reduction in Lavandula latifolia (Labiatae): A test of alternative hypotheses. Evolutionary Trends in Plants, 4, 99e105. Retrieved from: http://www.plant-animal.es/pdfs/Herrera.1990.Evol.Trends.Plants.pdf. Hofstede, F. E., & Sommeijer, M. J. (2006). Effect of food availability on individual foraging specialisation in the stingless bee Plebeia tobagoensis (Hymenoptera, Meliponini). Apidologie, 37(3), 387e397. http://dx.doi.org/10.1051/apido:2006009. Jers akov a, J., Johnson, S. D., & Kindlmann, P. (2006). Mechanisms and evolution of deceptive pollination in orchids. Biological Reviews, 81(2), 219e235. http:// dx.doi.org/10.1017/S1464793105006986. Knudsen, J., & Olesen, J. (1993). Buzz-pollination and patterns in sexual traits in North European Pyrolaceae. American Journal of Botany, 80(8), 900e913. Retrieved from: http://www.jstor.org/stable/2445510. Langenberger, M. W., & Davis, A. R. (2002). Temporal changes in floral nectar production, reabsorption, and composition associated with dichogamy in annual caraway (Carum carvi; Apiaceae). American Journal of Botany, 89(10), 1588e1598. http://dx.doi.org/10.3732/ajb.89.10.1588. Leonard, A. S., Dornhaus, A., & Papaj, D. (2011a). Flowers help bees cope with uncertainty: Signal detection and the function of floral complexity. Journal of Experimental Biology, 214, 113e121. http://dx.doi.org/10.1242/jeb.047407. Leonard, A. S., Dornhaus, A., & Papaj, D. (2011b). Forget-me-not: Complex floral displays, inter-signal interactions, and pollinator cognition. Current Zoology, 57, 215e224. Leonard, A. S., & Masek, P. (2014). Multisensory integration of colors and scents: Insights from bees and flowers. Journal of Comparative Physiology A, 200(6), 463e474. http://dx.doi.org/10.1007/s00359-014-0904-4. Lewis, A. (1993). Learning and the evolution of resources: Pollinators and flower morphology. In D. R. Papaj, & A. Lewis (Eds.), Insect learning: Ecological and evolutionary perspectives (pp. 219e242). New York, NY: Chapman & Hall. Lunau, K., Piorek, V., Krohn, O., & Pacini, E. (2015). Just spines: mechanical defense of malvaceous pollen against collection by corbiculate bees. Apidologie, 46(2), 144e149. http://dx.doi.org/10.1007/s13592-014-0310-5. Mayntz, D. (2005). Nutrient-specific foraging in invertebrate predators. Science, 307(5706), 111e113. http://dx.doi.org/10.1126/science.1105493. Menzel, R. (2012). The honeybee as a model for understanding the basis of cognition. Nature Reviews Neuroscience, 13(11), 758e768. http://dx.doi.org/10.1038/nrn3357. Menzel, R., & Giurfa, M. (2001). Cognitive architecture of a mini-brain: The honeybee. Trends in Cognitive Sciences, 5, 62e71. Monsell, S. (2003). Task switching. Trends in Cognitive Sciences, 7(3), 134e140. http://dx.doi.org/10.1016/S1364-6613(03)00028-7. Mosquin, T. (1971). Competition for pollinators as a stimulus for the evolution of flowering time. Oikos, 22(3), 398. http://dx.doi.org/10.2307/3543864.

132

F. Muth et al. / Animal Behaviour 126 (2017) 123e133

Muth, F., Francis, J. S., & Leonard, A. S. (2016). Bees use the taste of pollen to determine which flowers to visit. Biology Letters, 12(7), 20160356. Muth, F., Keasar, T., & Dornhaus, A. (2015). Trading off short-term costs for long-term gains: How do bumblebees decide to learn morphologically complex flowers? Animal Behaviour, 101, 191e199. http://dx.doi.org/10.1016/j.anbehav.2014.12.024. Muth, F., Papaj, D. R., & Leonard, A. S. (2015). Colour learning when foraging for nectar and pollen: Bees learn two colours at once. Biology Letters, 11(9). http:// dx.doi.org/10.1098/rsbl.2015.0628, 20150628. Muth, F., Papaj, D. R., & Leonard, A. S. (2016). Bees remember flowers for more than one reason: Pollen mediates associative learning. Animal Behaviour, 111, 93e100. http://dx.doi.org/10.1016/j.anbehav.2015.09.029. Nepi, M., Guarnieri, M., & Pacini, E. (2001). Nectar secretion, reabsorption, and sugar composition in male and female flowers of Cucurbita pepo. International Journal of Plant Sciences, 162(2), 353e358. http://dx.doi.org/10.1086/319581. Nicholls, E., & Hempel de Ibarra, N. (2014). Bees associate colour cues with differences in pollen rewards. Journal of Experimental Biology, 217(15), 2783e2788. http://dx.doi.org/10.1242/jeb.106120. Nicolson, S. W. (2011). Bee food: The chemistry and nutritional value of nectar, pollen and mixtures of the two. African Zoology, 46, 197e204. http://dx.doi.org/ 10.3377/004.046.0201. Nisenbaum, R. A., Patselas, M. G., & Weiner, S. D. (1999). Does flower color change in Aster vimineus cue pollinators? American Midland Naturalist, 141(1), 59e68. http://dx.doi.org/10.1674/0003-0031(1999)141[0059:DFCCIA]2.0.CO;2. O'Donnell, S., Reichardt, M., & Foster, R. (2000). Individual and colony factors in bumble bee division of labor (Bombus bifarius nearcticus Handl; Hymenoptera, Apidae). Insectes Sociaux, 47(2), 164e170. http://dx.doi.org/10.1007/PL00001696. Page, R., Scheiner, R., Erber, J., & Amdam, G. (2006). The development and evolution of division of labor and foraging specialization in a social insect (Apis mellifera L.). Current Topics in Developmental Biology, 74, 253e286. http://dx.doi.org/ 10.1016/S0070-2153(06)74008-X. Pashler, H. (1994). Dual-task interference in simple tasks: Data and theory. Psychological Bulletin, 116, 220e244. Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D., & R Core Team. (2016). nlme: Linear and nonlinear mixed effects models (R package version 3.1-128). Vienna, Austria: R Foundation for Statistical Computing http://CRAN.R-project.org/package¼nlme. Pleasants, J. M., & Chaplin, S. J. (1983). Nectar production rates of Asclepias quadrifolia: Causes and consequences of individual variation. Oecologia, 59(2e3), 232e238. http://dx.doi.org/10.1007/BF00378842. Pohl, M., Watolla, T., & Lunau, K. (2008). Anther-mimicking floral guides exploit a conflict between innate preference and learning in bumblebees (Bombus terrestris). Behavioral Ecology and Sociobiology, 63, 295e302. http://dx.doi.org/ 10.1007/s00265-008-0661. Raguso, R. A. (2004). Flowers as sensory billboards: Progress towards an integrated understanding of floral advertisement. Current Opinion in Plant Biology, 7(4), 434e440. http://dx.doi.org/10.1016/j.pbi.2004.05.010. Raine, N. E., & Chittka, L. (2007). Pollen foraging: Learning a complex motor skill by bumblebees (Bombus terrrestris). Naturwissenschaften, 94(6), 459e464. http:// dx.doi.org/10.1007/s00114-006-0184-0. Raine, N. E., & Chittka, L. (2008). The correlation of learning speed and natural foraging success in bumble-bees. Proceedings of the Royal Society B: Biological Sciences, 275(1636), 803e808. http://dx.doi.org/10.1098/rspb.2007.1652. Raubenheimer, D., & Tucker, D. (1997). Associative learning by locusts: Pairing of visual cues with consumption of protein and carbohydrate. Animal Behaviour, 54, 1449e1459. http://dx.doi.org/10.1006/anbe.1997.0542. Renner, S. (2006). Rewardless flowers in the angiosperms and the role of insect cognition in their evolution. In N. Waser, & J. Ollerton (Eds.), Plantepollinator interactions: From specialization to generalization (pp. 123e144). Chicago, IL: University of Chicago Press. Ruedenauer, F. A., Spaethe, J., & Leonhardt, S. D. (2015). How to know which food is good for you: Bumblebees use taste to discriminate between different

concentrations of food differing in nutrient content. Journal of Experimental Biology, 218(14), 2233e2240. http://dx.doi.org/10.1242/jeb.118554. Russell, A. L., Golden, R. E., Leonard, A. S., & Papaj, D. R. (2015). Bees learn preferences for plant species that offer only pollen as a reward. Behavioral Ecology, 27, 731e740. http://dx.doi.org/10.1093/beheco/arv213. Russell, A., & Papaj, D. (2016). Artificial pollen dispensing flowers and feeders for bee behaviour experiments. Journal of Pollination Ecology, 18, 13e22. Shettleworth, S. J. (2010). Cognition, evolution, and behavior (2nd ed.). Oxford, U.K.: Oxford University Press. Simpson, S., & Raubenheimer, D. (2012). The nature of nutrition: A unifying framework from animal adaptation to human obesity. Princeton, NJ: Princeton University Press. Simpson, S. J., Sibly, R. M., Lee, K. P., Behmer, S. T., & Raubenheimer, D. (2004). Optimal foraging when regulating intake of multiple nutrients. Animal Behaviour, 68, 1299e1311. http://dx.doi.org/10.1016/j.anbehav.2004.03.003. Skorupski, P., & Chittka, L. (2010). Photoreceptor spectral sensitivity in the bumblebee, Bombus impatiens (Hymenoptera: Apidae). PLoS One, 5, e12049. Sulikowski, D., & Burke, D. (2007). Food-specific spatial memory biases in an omnivorous bird. Biology Letters, 3(3), 245e248. http://dx.doi.org/10.1098/ rsbl.2007.0122. Sun, H., Huang, B., Yu, X., & Kou, Y. (2011). Reproductive isolation and pollination success of rewarding Galearis diantha and non-rewarding Ponerorchis chusua (Orchidaceae). Annals of Botany, 107(1), 39e47. http://dx.doi.org/10.1093/aob/ mcq250. Thomson, J. D. (1986). Pollen transport and deposition by bumble bees in Erythronium: Influences of floral nectar and bee grooming. Journal of Ecology, 74(2), 329. http://dx.doi.org/10.2307/2260258. Thomson, J., McKenna, M., & Cruzan, M. (1989). Temporal patterns of nectar and pollen production in Aralia hispida: Implications for reproductive success. Ecology, 70, 1061e1068. Vaudo, A. D., Patch, H. M., Mortensen, D. A., Tooker, J. F., & Grozinger, C. M. (2016). Macronutrient ratios in pollen shape bumble bee (Bombus impatiens) foraging strategies and floral preferences. Proceedings of the National Academy of Sciences of the United States of America, 113(28), E4035eE4042. http://dx.doi.org/ 10.1073/pnas.1606101113. Velthuis, H. H. W., & van Doorn, A. (2006). A century of advances in bumblebee domestication and the economic and environmental aspects of its commercialization for pollination. Apidologie, 37(4), 421e451. http://dx.doi.org/10.1051/ apido:2006019. Vergoz, V., Roussel, E., Sandoz, J.-C., & Giurfa, M. (2007). Aversive learning in honeybees revealed by the olfactory conditioning of the sting extension reflex. PLoS One, 2(3), e288. http://dx.doi.org/10.1371/journal.pone.0000288. Waldron, E. M., & Ashby, F. G. (2001). The effects of concurrent task interference on category learning: Evidence for multiple category learning systems. Psychonomic Bulletin & Review, 8(1), 168e176. http://dx.doi.org/10.3758/ BF03196154. Waser, N. (1983). The adaptive nature of floral traits: Ideas and evidence. In L. Real (Ed.), Pollination biology (pp. 241e285). Orlando, FL: Academic Press. Waser, N. (1986). Flower constancy: Definition, cause, and measurement. American Naturalist, 127, 593e603. Weiss, M. R., & Papaj, D. R. (2003). Colour learning in two behavioural contexts: How much can a butterfly keep in mind? Animal Behaviour, 65, 425e434. http:// dx.doi.org/10.1006/anbe.2003.2084. Worden, B. D., Skemp, A. K., & Papaj, D. R. (2005). Learning in two contexts: The effects of interference and body size in bumblebees. Journal of Experimental Biology, 208(11), 2045e2053. http://dx.doi.org/10.1242/jeb.01582.

Appendix

Table A1 The number of bees included in each treatment Experiment

Treatment

Colour trained to

No. of bees trained and tested

No. of bees included in final analysis*

1 1 1 1 2a 2a 2a 2a 2b 2b 2b

1: 1: 2: 2: 1: 1: 2: 2: 1: 2: 3:

Blue Yellow Blue Yellow Blue Yellow Blue Yellow Yellow Yellow Yellow

15 15 21 20 15 15 21 23 18 22 23

13 15 14 15 15 15 11 13 16 13 11

Nectar only Nectar only Nectar and pollen (on all flowers) Nectar and pollen (on all flowers) Pollen only Pollen only Pollen and nectar (on all flowers) Pollen and nectar (on all flowers) Pollen only Pollen and nectar on the same flower (yellow) Pollen on one flower (yellow), nectar on the other (blue)

Refer to Fig. 1 for a diagram of the experiments and treatments. * Fewer bees were included in the final analysis than went through training and testing because bees were excluded if they only collected a single reward in multireward treatments, or if they did not visit enough flowers (for details, see Data Analysis).

F. Muth et al. / Animal Behaviour 126 (2017) 123e133

133

blue

ree

e-g

UV -bl ue

blu n

b g y

V

gre

U

en

Figure A2. Topedown view of training array of artificial flowers (10 blue, 10 yellow). A bee is shown collecting pollen from the top-left yellow flower's anther.

UV-green

Figure A1. The yellow (‘y’) and blue (‘b’) flower colour cues that bees were trained to in the current experiments and the grey (‘g’) pretraining colour cue. Centre represents the background. Colours are plotted into bee colour space (Chittka, 1992) taking into account the photoreceptor spectral sensitivities of B. impatiens (Skorupski & Chittka, 2010), using methods as in Muth, Papaj, et al. (2015, 2016).